Introduction: Locally recurrent prostate cancer after primary radiotherapy might be curatively treated using focal salvage. Optimal patient selection is essential. This study therefore aims to develop a prediction model for biochemical failure (BF) following focal salvage high intensity focused ultrasound (HIFU).

Materials and methods: A comprehensive registry was used (Nov 2006-Sept 2014). Recurrences were assessed with multi-parametric MRI combined with template prostate mapping biopsies or systematic TRUS-guided biopsies as well as PET/CT and a bone-scan. Quadrant or hemi-ablation was applied. With Cox proportional-hazards regression, the effect of determinants on BF (Phoenix-definition) was assessed. Multiple imputation was used for missing data. Internal validation was performed using bootstrap resampling (500 datasets). Calibration plots and the C-statistic were adopted to assess model performance. Finally, a risk score was created.

Results: 139 patients were identified. Primary treatment consisted of external beam radiotherapy (EBRT, n=134) or EBRT with high dose rate brachytherapy (n=5). Mean follow-up was 37 months (sd 21 months). BF occurred in 71, resulting in 50% biochemical disease-free survival (bDFS) at 35 months. Multivariable, disease-free survival interval (DFSI) after primary radiotherapy, pre-salvage PSA, PSA doubling time (PSADT), prostatic volume and T-stage (both MRI-based) were independent predictors (Table 1). The adjusted C-statistic was 0.69. Calibration was accurate up to 36 months. Our risk score consisted of 4 groups, highly predictive of bDFS at 3 years (88% [95%CI: 75-100%], 68% [51-81%], 33% [18-48%] and 20% [0-40%].

Conclusion: Improved focal salvage patient selection can increase recurrence free rates from 50% to almost 90% at 3 years.